Back

Bivariate GSA-MiXeR: A Novel Tool for Functional Genomic Analyses Implicates Diverse Neural Cell Types for Psychiatric and Neurodegenerative Disorders

Parker, N.; Furher, J.; Nguyen, D.; Fominykh, V.; Jaholkowski, P.; Akkouh, I.; O'Connell, K. S.; Hagen, E.; Bahrahmi, S.; Arsland, D.; Bergh, S.; Engstad, T.; Fladby, T.; Knapskog, A.-B.; Persson, K.; Grontvedt, G. R.; Madsen, B.-O.; Rongve, A.; Saltvedt, I.; Sando, S. B.; Scheffler, K.; Selbaek, G.; Stordal, E.; Toft, M.; Watne, L. O.; Djurovic, S.; Smeland, O. B.; Dale, A. M.; Shadrin, A. A.; Andreassen, O. A.; Frei, O.

2025-12-17 genetic and genomic medicine
10.64898/2025.12.16.25342384 medRxiv
Show abstract

The growing number of genomic discoveries in complex human traits has highlighted the need for advanced functional genomics tools that parse their polygenic and pleiotropic genetic architecture to provide biological insights. We present bivariate GSA-MiXeR, a novel tool that models the partitioned heritability and covariance of two traits within a genomic region of interest (ROI) and estimates the (i) trait-specific fold enrichment, (ii) local genetic correlation, and (iii) local genetic omnibus statistic for ranking genomic ROIs. Unlike previous methods, our tool estimates local genetic correlations both in continuous and disjoint genomic ROIs, expanding the ability to assess local genetic overlap among complex traits. We perform simulations to validate our tool and illustrate its utility in applied analyses that implicate diverse neural cell types for psychiatric and neurodegenerative disorders using single cell RNA sequencing data. Bivariate GSA-MiXeR provides new analytical avenues that facilitate a transition from genetic discovery to mechanistic insights. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=113 SRC="FIGDIR/small/25342384v1_ufig1.gif" ALT="Figure 1"> View larger version (45K): org.highwire.dtl.DTLVardef@5e60f2org.highwire.dtl.DTLVardef@2ed1e8org.highwire.dtl.DTLVardef@1d6f84aorg.highwire.dtl.DTLVardef@46ca65_HPS_FORMAT_FIGEXP M_FIG C_FIG

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.